Is it possible to modify things so a user can pass the output format at runtime?

At the moment it's hardcoded to: --output-srt

But whisper is capable of outputting as txt, vtt, lrc, csv, json, etc.

Not sure if it's easier to have this as a separate flag, or a different command for each output type.

Heya,

Just tried to use it to transcribe a file and it failed to download a model.

I set the model in the config to small and got Invalid username or password. in the file ggml-small.bin.

Update: changed the url in download_model to "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/${model_file}" and now it seems to work.